High end needs SCSI and maybe a Pentium Pro.

But 1995/1996 is more like 75-166. Pentium 200 MMX is from 1997.

http://www.cpu-world.com/CPUs/Pentium/Intel-P … FV80503200.html

Socket 8 full stuff hard to find and it was the Server CPU, not for Consumer/Gamers

Roman78 wrote:

High end needs SCSI and maybe a Pentium Pro.

Not for a machine running mostly DOS/Windows 95 games.

Pretty similar to my machine, so thumbs up at the choice of components! 🤣

A Voodoo 1 would be a good match.

CF adapter for easy data transfer or additional storage (the 4 GB drive will be too small before long).

Maybe a NIC for data transfer.

The Yamaha card has real OPL3, that's certainly a plus. Maybe add a MIDI daughterboard?

If you don't have a strong preference for OPL3, I'd go with the AWE64 instead. Mostly trouble free, sounds good, low noise. And CQM isn't THAT bad IMO. Some Creative cards with real OPL3 manage to sound worse (I'm looking at you, CT3980 😒 )

If you want the machine only for DOS games up to 1996. I'd limit to up to 1995 and get 486, not Pentium because of possible compatibility issues. For later games I'd use PII.

what

a p200mmx is fine for a lot of games, even some 1999 ones.
High-end 486s are often plagued by poor motherboard chipset issues and other limitations that highly vary that would also make it a headache for most 1996 games.

eFatal2ty wrote:

I want to build an High-End 1995-1996 Build

How strictly did you want to keep to that? You've got more of a 1997 build going with the 430TX and the 200mmx. Though I guess having "next year's computer" would be the High-End we always wanted.

The MMX is probably overkill for 1996 (it is a processor from 1997), in 1996 the fastest thing you could get was a Pentium 200. It will be powerful enough for everything from 1996.

I disagree.

Going overkill would be Pentium Pro, which was also available in 1996.

nothing in 1996 can make use of MMX, and MMX is overhyped trashy extensions that only really benefits audio processing in later 98+ games and console emulators

MMX has more L1 cache, so it is faster than a regular Pentium at the same clock speed. Also, PPro is relatively slow for 16-bit code. The MMX is a better fit here, while a Pentium Pro would be a much better period correct Windows NT machine.

leileilol wrote:

a p200mmx is fine for a lot of games

486 are more compatible with DOS games as some don't like Pentium's cache. Besides more possible speed issues.

Such as..........?

alexanrs wrote:

MMX has more L1 cache, so it is faster than a regular Pentium at the same clock speed. Also, PPro is relatively slow for 16-bit code. The MMX is a better fit here, while a Pentium Pro would be a much better period correct Windows NT machine.

This is often forgotten. I believe a 166 MMX is quite bit faster than a Pentium 200. The MMX also runs on lower voltage, draws less power, produces less heat. The 233 MMX makes a very capable retro PC.

philscomputerlab wrote:

I believe a 166 MMX is quite bit faster than a Pentium 200.

I seriously doubt that in pure DOS applications the core enhancements of the P55C will make up for the difference in operating frequency against the P54C. At the same operating frequency the MMX will be probably around 5% better, this should be valid for the majority of legacy DOS applications.

In Windows your statement probably nails it.

On my ex Pentium mmx 233 I could run almost everything fine, if there was some dos game that wasn't running I would just disable l1 or l2 cache or even run moslo 😁

At the end, I ended up to a k6-iii+, for being able to change the clock by software, so I could have a more precise control about performance over old games (and also I could play better games from 1998+

carlostex wrote:

philscomputerlab wrote:

I believe a 166 MMX is quite bit faster than a Pentium 200.

I seriously doubt that in pure DOS applications the core enhancements of the P55C will make up for the difference in operating frequency against the P54C. At the same operating frequency the MMX will be probably around 5% better, this should be valid for the majority of legacy DOS applications.

In Windows your statement probably nails it.

Yes, in my Voodoo 2 (and SLI) scaling project, the MMX 166 beats the Pentium 200 easily. Under DOS I also found the MMX 166 to be faster. There is data in the VGA Benchmark database if you're keen.

philscomputerlab wrote:

Yes, in my Voodoo 2 (and SLI) scaling project, the MMX 166 beats the Pentium 200 easily.

What did i say?

philscomputerlab wrote:

Under DOS I also found the MMX 166 to be faster. There is data in the VGA Benchmark database if you're keen.

You can't be serious. Where and how is the VGA Benchmark a good source to compare a P54C @200 and a P55C @166?

The Ultimate 686 Benchmark Comparison

http://www.tomshardware.co.uk/does-the-pentiu … ,review-19.html

The Pentium MMX is indeed internally more advanced than the classic Pentium and applications that benefit from bigger caches and bigger cache associativity will be faster on the MMX. However the MMX will suffer with higher penalty in pipelines stalls because it has deeper pipeline stages than the classic Pentium. The problem with comparing CPU's is that there are tons of variables that have to be considered. The Pentium 166 MMX will hardly be always faster than the classic Pentium 200. My biggest problem was with your wording:

"I believe a 166 MMX is quite bit faster than a Pentium 200."

How much is quite a bit? Quite a bit faster sounds significantly faster. And just like i said true for Windows but not so much for DOS.

ibm5155 wrote:

At the end, I ended up to a k6-iii+, for being able to change the clock by software, so I could have a more precise control about performance over old games (and also I could play better games from 1998+

I've found that AMD-K6's are a bit sensitive to the Turbo Pascal bug. FIFA94 also gives me a memory segmentation fault that doesn't happen with Pentium CPU's, either MMX or classic. However the K6+ ability to clock down and up via software is an ability that has become hard to ignore. 😀

What did i say?

I was agreeing with you.

You can't be serious. Where and how is the VGA Benchmark a good source to compare a P54C @200 and a P55C @166?

The whole point of the VGA benchmark project was to collect lots of data to answer questions such as this 😐

Did your benchmarking find that DOS benchmarks are slower on the MMX 166 compared to the Pentium 200?

philscomputerlab wrote:

The whole point of the VGA benchmark project was to collect lots of data to answer questions such as this 😐

If you're trying to answer specific questions with a general approach you are bound to fail and answer nothing. I appreciate all your efforts and work you put into your projects and benchmarks, but when you suggest the VGA bench database to compare 2 different CPU's in a very specific way it's like i'm communicating with a different Phil.

The VGA benchmarh database is great to compare overall systems with different hardware on them which data is supplied by different users. How many different 386DX-40 systems are in the database? Which motherboard and VGA card did the fastest 386 system use? How much faster was it let's say comparing with the guy that used different VGA and motherboard but used more agressive memory timings or faster cache? And BTW, does this benchmark database show that the fastest 386 system will be ALWAYS faster than other similar systems in ALL similar DOS applications?

philscomputerlab wrote:

Did your benchmarking find that DOS benchmarks are slower on the MMX 166 compared to the Pentium 200?

I haven't done any benchmarking to specifically compare these 2 CPU's if that's what you want to know. This is for several reasons. The most important ones is that the CPU's are well documented to understand the differences between the 2 and there is enough information today to answer these questions. Then i also take in consideration how inneficient can benchmarks (or should i say benchmarketing?) be to tell the real performance or compare CPU's.

You may think that i'm being all so arrogant because i don't provide any of my benchmarks, sorry but i'm not benchmark guy, and that's simply because i find benchmarking to only tell a small part of a much bigger picture. If i ever come by a Pentium 200 reasonably priced i'll indulge you and test against my Pentium 233 MMX that i luckily got bundled with my ASUS TX-97XE. I also don't want to spend my retro computing time on an issue that's been covered for years and where i won't uncover any secrets. I'm much more interested in working in projects such as expanding the retro capabilities of my build, via better motherboard capabilities, hardware mods, sound card replica projects and so on.

I gave a link to feipoa's thread because it gives a bigger picture on how in a bigger window of applications the P55C is indeed faster than the classic Pentium on a clock per clock basis. I'm merely contesting the " MMX is quite a bit faster than the classic" argument in which you include DOS. So far your benchmarking only showed that the MMX is the better choice when using Voodoo cards, which is hardly surprising. Can't you realize how vague this statement is?

I understand your concerns.

I've done my own benchmarks ages ago, I will see if I can find them when I get home.

With the database, often users have used the same system, but changed the processor. With the filter options in Excel that should be easy to find. I'll have a look. You can only select Pentium and MMX processors for example, then only 166 and 200 models, then only selected users...

I just assumed that people use this function. If course, comparing other systems with different specifications is not going to give you a real answer.

EDIT: Turns out nobody submitted results for a Pentium 200, so that's unfortunate.

EDIT EDIT: Ok I looked at the data I collected. It was in 2013, unreal how time flies past 😵 . I remembered wrongly, under DOS, at least 3dbench2, the Pentium 200 is in front of the MMX 166.

I've attached the benchmark results if anyone is interested.